Multiscale voter model on real networks

We introduce the Multiscale Voter Model (MVM) to investigate clan influence at multiple scale -- family, neighborhood, political party... in opinion formation on real complex networks. Clans, consisting of similar nodes, are constructed using a coarse-graining procedure network embeddings that allows us control for length interactions. ran numerical simulations monitor evolution MVM dynamics and synthetic networks, identified transition between final stage full consensus one with mixed binary opinions. The depends clans strength their influence. found enhancing group diversity promotes while strong kinship yields metastable clusters same opinion. segregated domains, which signal polarization, discernible as spatial patterns hyperbolic Our multiscale framework can be easily applied other dynamical processes affected by